1. Field of the Invention
The invention relates to transmission of unprocessed biometric data from a camera or other sensor to a server at a remote location over a network in a secure manner.
2. Background of the Invention
Many situations occur in which it is necessary to identify a person seeking to enter a secured location, use a computer system or perform a financial or other transaction to assure that the person is authorized to perform the task. There are several methods known as biometrics for recognizing or identifying an individual. These methods include analyzing a signature, obtaining and analyzing an image of a fingerprint and imaging and analyzing the retinal vascular patterns of a human eye. Recently the art has used the iris of the eye which contains a highly detailed pattern that is unique for each individual and stable over many years as a non-contact, non-obtrusive biometric. This technique is described in U.S. Pat. Nos. 4,641,349 to Flom et al. and 5,291,560 to Daugman. Biometric identification systems take an image of the person being identified at the time he seeks to perform the task. This image is then processed to extract certain features. The result of this processing is an iris code in the case of U.S. Pat. No. 5,291,560 or in more general terms, a biometric template. This newly computed biometric template is then compared with a previously filed biometric template to make the identification and then authorize or refuse the person to perform the task.
Verification of an individual's identity or identification of an individual via automated biometric devices such as iris systems and some fingerprint systems rely on digital imaging technologies. The raw biometric data of the individual's presumably unique features is obtained by a system consisting of optics, camera, and electronics which capture and digitize the scene presented to the camera. The digital representation of the image (i.e. the raw or unprocessed image data) is then processed by an algorithm which converts the image data into a particular representation, called a biometric template. The biometric template is suitable for matching against a previously stored template to verify the individual's identity or against multiple templates to identify an individual. This method is illustrated in the flow chart shown in FIG. 1. Frequently, the conversion to the biometric template and subsequent matching is performed by a computer that is situated at a location remote from the camera or sensor which collects the biometric data. This remote conversion and matching is done so that the integrity of the biometric template computation algorithm is maintained by not distributing it to non-secure locations such as in a user's home computer. Thus, the digitized unprotected biometric data must then be transmitted from the camera to the remote computer. Such transmission may occur over a direct and dedicated transmission line, over a public carrier such as a telephone system or even over the Internet.
Any biometric system can be defrauded if an attacker could substitute the raw image data prior to the conversion to a biometric template and subsequent matching. That is to say, if Mallory wished to impersonate Bob, he would first capture Bob's raw image data and store it; then he could attack the target system by artificially "injecting" the fraudulent image data at the correct time and place so that the template conversion and match would unwittingly conclude that it was Bob appearing in front of the camera. The potential for such an attack increases when the image acquisition process may be separated from the template conversion and match process, such as in a case where a remote terminal equipped with the imaging system (a client) sends images across an internal network, or the Internet, to a central server which performs the conversion and match. Image tampering can occur at any point from the camera to a "secured" server system. The server system itself may also be attacked, but the probability of a successful attack against this system is extremely small without "inside" information and access.
There are several key places where an attacker could perform this image substitution. An attacker could replace the camera with a system that mimics the camera functionality thereby providing a previously stored image to the rest of the system. An attacker could gain access to the inside of the client system's host and replace the contents of the memory or frame store containing the "real" image data with the memory representation of a previously stored image. Finally, the attacker could gain access somewhere along the communications path between the client system and the server system and replace the image while in transit. Thus, there is a need for a method and device which can transmit biometric data while preventing image substitution or tampering.
We recognize that attacks could be made using an artificial or contrived scene. For example, an attacker could present a counterfeit eye to an otherwise unmolested system. These are completely different types of attacks. The technology to counter such threats exists and attacks of this type are not relevant to the present discussion. Rather the present invention is aimed at maintaining the integrity of an image containing biometric data and preventing image tampering or substitution.
There have been developed a number of techniques for detecting image or data tampering and unauthorized copying. Much of this effort has been directed to preventing and detecting copyright infringement and counterfeiting. The most widely used techniques apply a watermark on the image or embed code signals. U.S. Pat. Nos. 5,768,426; 5,809,139 and 5,822,432 disclose methods for marking digital video signals by adding bits of information to the signal or image file in a predetermined way so that the data appears as noise to the ordinary observer but can be detected as a watermark or code by the owner of the signal or image file. U.S. Pat. No. 5,613,004 discloses a steganographic method and device that encodes a digitized data stream with special keys. The patent also teaches that codes or other information can be prepended or appended to the data stream. Another known technique for watermarking images is to change the brightness of selected pixels in a predetermined pattern. This method is disclosed in U.S. Pat. No. 5,825,892. However, none of these references are concerned with assuring that biometric data has not been compromised to prevent unauthorized access to a secure system or location.
It is imperative that raw biometric image data be secured in such a manner that undetectable substitution or tampering of the image data prior to the biometric template conversion is extremely difficult to perform. In addition, it may be desirable to encode image data such that images have a finite lifetime. Then an encoded image cannot be used to identify a user more than once (or n times), and/or images remain valid only for some predetermined time period after which they will not be processed by the biometric algorithm since they have been deemed invalid by the authentication system or server. Furthermore, it may be desirable that the biometric authentication server be provided with a unique ID of the imaging system providing the image data to it. With these features, an authentication server has the ability to determine that it is indeed Bob in front of camera ID#xyz for transaction ID#pdq, that capture occurred within time interval t2-t1, and that the image could not have been altered nor been re-used from some other transaction.
For example, in an electronic commerce application, for each transaction there would be exactly one image associated with that transaction. Also, should the client fail to provide the server with the image within some time window, the transaction would not be authorized. This provides additional levels of protection by preventing later substitution of previously valid images and forcing an attacker to develop methods that work within the time-out period, which adds considerable difficulty to the attacker's task.
In order to provide adequate security, there must be a "secret" shared only by the sender (the camera) and the recipient (the system performing the authentication of the image.) This secret must remain secure, or else the security of the entire system may be compromised. The "secret" shared between the camera and the authentication server is in the form of a digital "key" or, in some cases, a pair of keys. The camera's tamper-resistant/responsive packaging protects the secret key embedded inside. The server's key is protected by security measures such as electronic firewalls, physical access control, and other known methods of providing high levels of security to sensitive data.
In recent years there has been an increase in the use of personal computers both in the office and at home. Initially these computers were used almost exclusively for word processing and database applications. Today personal computers are being used for a variety of communications activities ranging from electronic mail and file transfer to electronic banking and online commerce. Thus, there is a need for a system that can be connected to a desktop computer which can ensure the secure transmission and reception of biometric data over an unsecured network thereby enabling a user to have his identity verified at a remote location so that such user can be authorized to make the financial transfer or transaction that he has requested.